Scholarlink

<p>Accurate monitoring of Leaf area index (LAI) is essential in precision agriculture. This study aimed to enhance the accuracy of rice LAI estimation by integrating Grey level co-occurrence matrix (GLCM) textural features, spectral bands (SB), color indices (CIs), vegetation indices (VIs), and plant height; and by employing machine learning (ML) models and Beer-Lambert Law. The UAV-based RGB and multispectral (MS) sensors were used to collect the images from two seasons (2023 and 2024) of rice grown in Roorkee, Uttarakhand, India. Five ML models were employed, Extreme Gradient Boosting (XGBoost), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Random Forest (RF), and Artificial Neural Network (ANN) with three feature selection methods: Least Absolute Shrinkage and Selection Operator (LAS), Maximal Information Coefficient (MIC), and Recursive Feature Elimination (RFE). A total of 75 input features were utilized. Among all the model combinations, XGBoost-RFE (R<sup>2</sup> = 0.86, NRMSE = 15%) had the highest prediction accuracy. The three combinations of the RF model (RF-MIC, RF-RFE, and RF-LAS) performed similarly (R<sup>2</sup> = 0.85, NRMSE = 16%), and their performance was comparable to that of XGBoost-RFE (R<sup>2</sup> = 0.86) and the Beer-Lambert law (R<sup>2</sup> = 0.83). The GLCM Dissimilarity Green textural feature highly influenced the rice LAI. The color indices, including EG, ER, WI, GLA, EXGR, and EB; spectral bands RedEdge, NIR, Green, Red, and Blue; and VIs comprising RVI, NDI, and ENDVI affected the rice LAI. Overall, UAV-based RGB and MS imagery coupled with ML models successfully predicted rice LAI, offering a reliable solution for precision crop monitoring.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Spatial mapping of rice leaf area index using UAV-borne RGB and multispectral data: a comparison of Beer-Lambert law and machine learning algorithms

  • 期刊论文

  • 发表日期：

    2026年5月19日

查看全文
  • Sumit Kumar Vishwakarma,
  • Kritika Kothari,
  • Ashish Pandey

摘要

Accurate monitoring of Leaf area index (LAI) is essential in precision agriculture. This study aimed to enhance the accuracy of rice LAI estimation by integrating Grey level co-occurrence matrix (GLCM) textural features, spectral bands (SB), color indices (CIs), vegetation indices (VIs), and plant height; and by employing machine learning (ML) models and Beer-Lambert Law. The UAV-based RGB and multispectral (MS) sensors were used to collect the images from two seasons (2023 and 2024) of rice grown in Roorkee, Uttarakhand, India. Five ML models were employed, Extreme Gradient Boosting (XGBoost), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Random Forest (RF), and Artificial Neural Network (ANN) with three feature selection methods: Least Absolute Shrinkage and Selection Operator (LAS), Maximal Information Coefficient (MIC), and Recursive Feature Elimination (RFE). A total of 75 input features were utilized. Among all the model combinations, XGBoost-RFE (R2 = 0.86, NRMSE = 15%) had the highest prediction accuracy. The three combinations of the RF model (RF-MIC, RF-RFE, and RF-LAS) performed similarly (R2 = 0.85, NRMSE = 16%), and their performance was comparable to that of XGBoost-RFE (R2 = 0.86) and the Beer-Lambert law (R2 = 0.83). The GLCM Dissimilarity Green textural feature highly influenced the rice LAI. The color indices, including EG, ER, WI, GLA, EXGR, and EB; spectral bands RedEdge, NIR, Green, Red, and Blue; and VIs comprising RVI, NDI, and ENDVI affected the rice LAI. Overall, UAV-based RGB and MS imagery coupled with ML models successfully predicted rice LAI, offering a reliable solution for precision crop monitoring.